The Application Of BP Neural Network Optimized By Genetic Algorithm In Rainfall Forecasting

In recent years, frequent floods occurred in Jiangxi Yichun C ity, almost every year there are different degrees o f flooding disaster. Especially in recent years, flood disasters occurred frequently and was exacerbated by the trend.C urrently,Yichun C ity Hydrology Bureau under the various counties,cities and districts have established a large number of Hydrology basic information collection site.However,there are still some deficiencies in annual rainfall prediction of the existing system and the system uses traditional linear regression methods. Rainfall contains a lot of uncertain and random factors, it is a time-varying non-stationary time series and therefore the existing system can not make a prediction to some extent on the future of Hydrology information based on existing data. Forecasting the rainfall in a region can help agriculture and water conservancy departments effectively improve the ability of preventing and controlling flood disasters, which will lead to the minimum, so the forecast of the rainfall becomes an urgent research task.It is very difficult to predict the rainfall by traditional methods. In recent years, artificial intelligence technology continues to mature prediction, BP neural network has been in rainfall prediction in a wide range of applications,with relatively good nonlinear learning ability,to some extent,be able to portray rainfall uncertainty and random variation,improved rainfall prediction accuracy. But with the BP neural network is widely used in many fields, many defects of basic BP algorithms including the slow convergence speed and local extremun gradually been discovered in the process of adjusting its weights and thresholds. We introduce genetic algorithm to their weights and thresholds be optimized to achieve a good training effect.Aiming at the deficiency of existing methods for rainfall forecasting,the BP neural network model is introduced to the rainfall forecast system of Hydrological Bureau in Yichun city. Taking rainfall and flood forecast information as the application background,based on Yichun C ity Bureau of hydrology,rainfall forecast status,at the same time,Yichun C ity,Jiangxi Province 1988 to 2010 years of annual rainfall is based. Respectively,using standard BP algorithm and has a global search to find excellent genetic algorithm optimization BP neural network model was constructed to predict rainfall in the region. Using Matlab neural network toolbox,calculated by the computer program and improving methods of analysis,verification of application performance improved BP neural network. The results show that the improved algorithm significantly improves the fitting acc uracy so that the model has a strong applicability and replicability.